Advanced Materials Research Vols. 403-408

Abstract: This study put forward container port competitiveness index evaluation system and a new method of division of hinterland basing on the evaluation system and a p-choice model. The result can provide reference for macro-plan and construction of port scale of Chinese container ports and avoid unnecessary waste that is caused by blind competition.

Abstract: Based on explaining characteristics of aluminum production process and management, this paper built a production scheduling model of the aluminum casting with its equipment constrains and process constraints, discussed the optimization strategy of the production scheduling for aluminum casting.

Abstract: This study aimed to the accuracy comparison of data imputation estimation methods between the unconstrained structural equation modeling (Uncon-SEM) and k-nearest neighbors (K-NN). The measurement accuracy of the model based on the mean magnitude of relative error (MMRE). The model is developed by using the online database from University of California, Irvine (UCI) which is a data set on waveform generators. Indicators 21 (1,200 sets) methods were as follows: 1) Data set was divided into two groups (experimental group of 1,000 sets and test group of 200 sets); 2) The experimental group was analyzed by three main factors (F1, F2, F3); 3) Uncon-SEM method: It created a SEM with three main factors, then the remaining factors to be created new the relationships with the unconstrained approach and created new SEM. The test data was substituted in the equation to find the MMRE which was 34.29% (accuracy was 65.71%); 4) K-NN method: It selected the main factor was the relationship of the missing data (F2). Measure the Euclidean distance between the test group and experimental group and selected 5 (K=5) of data sets were nearest to the missing data for the estimate by mean. The MMRE which was 57.00% (accuracy was 43.00%). Thus, comparing estimates of missing data showed that using the Uncon-SEM method were more accuracy, and MMRE declined about 22.71% than K-NN method.

Abstract: Based on Multi-Objective Memetic Algorithm (MOMA), a novel Multi-Objective Chaos Memetic Algorithm (MOCMA) is proposed . MOCMA is presented to keep population’s diversity, avoid local optimum and improve performance of Multi-Objective Memetic Algorithm. By virtue of the over-spread character of chaos sequence, it is used to generate chromosome to overcome redundancies. At the same time, searching space is enlarged by using sensitivity of chaos initial value. The comparisons of MOCMA with NSGAII in DTLZ problems suggest that MOCMA clearly outperforms in converging towards the true pareto front and finding the spread of solutions.

Abstract: In this research we propose an ensemble classification technique based on decision tree, artificial neural network, and support vector machine models weighting classifier by adaboost in order to increase classification accuracy. we used a total of 30 classifiers. The technique generated random data used Bootstrap. Testing Diabites Data from UCI, classification accuracy tests on Diabites Data found that the proposed ensemble classification models weighting classifier by Adaboost yields better performance than that of a single model with the same type of classifier. The result as follows, Diabites Data achieved the best performance with 75.21%. we can conclude that there are two essential requirements in the model. The first is that the ensemble members or learning agents must be diverse or complementary, i.e., agents must exhibit different properties. Another condition is that an optimal ensemble strategy is also required to fuse a set of diverse by AdaBoost.

Abstract: Twitter has rapidly increased in popularity over the past few years. So, we have focused on Twitter as it has a large scale of data which is increasingly difficult to search through. In this paper, we propose recommendations for content on Twitter. We explored four dimensions in designing such as: topic relevance of content sources, the content candidate set for users, social voting and Meta data mapping. We implemented 24 algorithms for analysis of 12,000 records for three domains as follows: entertainment, stock exchange and smart phone in the design space. The best performing algorithm improved the percentage of correct matching interesting content to 23.86%.

Abstract: This paper proposes a prediction model for the PM₁₀ forecasting in Bangkok. Particulate matter (PM₁₀) with aerodynamic diameter up to 10 m (PM10) is targeted because these small particles effects people’s health and it constitutes major con_{Subscript text}cern for the air quality of Bangkok. Support vector regression (SVR) has been successfully employed to solve regression problem of nonlinearity. The determination for hyper-parameters including kernel parameters and the regularization is important to the performance of SVR. Particle swarm optimization (PSO) is a method for finding a solution of stochastic global optimizer based on swarm intelligence. Using the interaction of particles, PSO searches the solution space intelligently and finds out the best one. Thus, the proposed forecasting model based on the global optimization of PSO and local accurate searching of SVR is applied to forecast PM10 in this paper. The results of this research show the practical prediction model of PM10 based on PSO-SVR is established with C = 5009, ε = 0.0011, σ = 0.1072. The mean squared error (MSE) of the prediction model using PSO-SVR is about 8.654610-11. Practical results indicate that the application of the PSO-SVR method to temperature forecasting of PM10 is feasible and effective. The results show that the model is effective and highly accurate in the forecasting of PM₁₀.

Abstract: A universal problem with text classification has a problem due to the high dimensionality of feature space, e.g. word frequency vectors. To overcome this problem, this paper proposed a feature selection which focuses on statistical pattern based on SVM Attribute. Experiments have shown that the determination of word importance may increase the speed of the classification algorithm and save their resource used significantly. The proposed method was studied by comparing classification performance among Decision Tree, Naïve Bayes, and Support Vector Machine. The results showed that Support Vector Machine was found to be the best algorithm with F-measure 93.6%. It is found that the feature selection can reduce dimensionality of data significantly.

Abstract: The objectives of this research were to find out the Structural equation modeling coefficient and other parameter estimation under unknown prior distribution and compare this new model’s coefficient accuracy with the former model on “Web based application maintenance cost estimation multi group modeling” [16]. This new coefficient and parameter were estimated with Bayesian analysis instead of Maximum likelihood estimation (ML). The input model used in Bayesian analysis was started from the ML-model result [16]. The new values were replaced into the former model then MMRE was detected from 30 (testing) completed software maintenance projects while 192 projects were used for SEM model training. The result of cross validation was about 44.08% for Bayesian analysis refined SEM model while the ML-SEM model was 47.58%.

Abstract: Education Surveillance System is designed for predicting the state of education based on form of alarm signal using Incremental Leaning based on Mahalanobis Distance (ILM). However, ILM need to define two crucial parameters (co-variance matrix and distance threshold) it is not only very difficult for determining by general user but also depend on dataset property. This research proposed GAILM algorithm based on Ordinary National Education Test (Bangkok) dataset for finding approximate parameter and predicting. The result of experiment is represent GAILM technique discovering proximate co-variance matrix (0.91) and distance threshold parameter (0.44) and also high accuracy rate as 90.91% and 92.07%, in the year 2007 to 2008 respectively. This result was higher than the accuracy rate of traditional technique by K-Means algorithm and Cobweb.